~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/lib/inflate.c

Version: ~ [ linux-5.4-rc7 ] ~ [ linux-5.3.11 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.84 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.154 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.201 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.201 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.77 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-3.9.11 ] ~ [ linux-3.8.13 ] ~ [ linux-3.7.10 ] ~ [ linux-3.6.11 ] ~ [ linux-3.5.7 ] ~ [ linux-3.4.113 ] ~ [ linux-3.3.8 ] ~ [ linux-3.2.102 ] ~ [ linux-3.1.10 ] ~ [ linux-3.0.101 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 // SPDX-License-Identifier: GPL-2.0
  2 #define DEBG(x)
  3 #define DEBG1(x)
  4 /* inflate.c -- Not copyrighted 1992 by Mark Adler
  5    version c10p1, 10 January 1993 */
  6 
  7 /* 
  8  * Adapted for booting Linux by Hannu Savolainen 1993
  9  * based on gzip-1.0.3 
 10  *
 11  * Nicolas Pitre <nico@fluxnic.net>, 1999/04/14 :
 12  *   Little mods for all variable to reside either into rodata or bss segments
 13  *   by marking constant variables with 'const' and initializing all the others
 14  *   at run-time only.  This allows for the kernel uncompressor to run
 15  *   directly from Flash or ROM memory on embedded systems.
 16  */
 17 
 18 /*
 19    Inflate deflated (PKZIP's method 8 compressed) data.  The compression
 20    method searches for as much of the current string of bytes (up to a
 21    length of 258) in the previous 32 K bytes.  If it doesn't find any
 22    matches (of at least length 3), it codes the next byte.  Otherwise, it
 23    codes the length of the matched string and its distance backwards from
 24    the current position.  There is a single Huffman code that codes both
 25    single bytes (called "literals") and match lengths.  A second Huffman
 26    code codes the distance information, which follows a length code.  Each
 27    length or distance code actually represents a base value and a number
 28    of "extra" (sometimes zero) bits to get to add to the base value.  At
 29    the end of each deflated block is a special end-of-block (EOB) literal/
 30    length code.  The decoding process is basically: get a literal/length
 31    code; if EOB then done; if a literal, emit the decoded byte; if a
 32    length then get the distance and emit the referred-to bytes from the
 33    sliding window of previously emitted data.
 34 
 35    There are (currently) three kinds of inflate blocks: stored, fixed, and
 36    dynamic.  The compressor deals with some chunk of data at a time, and
 37    decides which method to use on a chunk-by-chunk basis.  A chunk might
 38    typically be 32 K or 64 K.  If the chunk is incompressible, then the
 39    "stored" method is used.  In this case, the bytes are simply stored as
 40    is, eight bits per byte, with none of the above coding.  The bytes are
 41    preceded by a count, since there is no longer an EOB code.
 42 
 43    If the data is compressible, then either the fixed or dynamic methods
 44    are used.  In the dynamic method, the compressed data is preceded by
 45    an encoding of the literal/length and distance Huffman codes that are
 46    to be used to decode this block.  The representation is itself Huffman
 47    coded, and so is preceded by a description of that code.  These code
 48    descriptions take up a little space, and so for small blocks, there is
 49    a predefined set of codes, called the fixed codes.  The fixed method is
 50    used if the block codes up smaller that way (usually for quite small
 51    chunks), otherwise the dynamic method is used.  In the latter case, the
 52    codes are customized to the probabilities in the current block, and so
 53    can code it much better than the pre-determined fixed codes.
 54  
 55    The Huffman codes themselves are decoded using a multi-level table
 56    lookup, in order to maximize the speed of decoding plus the speed of
 57    building the decoding tables.  See the comments below that precede the
 58    lbits and dbits tuning parameters.
 59  */
 60 
 61 
 62 /*
 63    Notes beyond the 1.93a appnote.txt:
 64 
 65    1. Distance pointers never point before the beginning of the output
 66       stream.
 67    2. Distance pointers can point back across blocks, up to 32k away.
 68    3. There is an implied maximum of 7 bits for the bit length table and
 69       15 bits for the actual data.
 70    4. If only one code exists, then it is encoded using one bit.  (Zero
 71       would be more efficient, but perhaps a little confusing.)  If two
 72       codes exist, they are coded using one bit each (0 and 1).
 73    5. There is no way of sending zero distance codes--a dummy must be
 74       sent if there are none.  (History: a pre 2.0 version of PKZIP would
 75       store blocks with no distance codes, but this was discovered to be
 76       too harsh a criterion.)  Valid only for 1.93a.  2.04c does allow
 77       zero distance codes, which is sent as one code of zero bits in
 78       length.
 79    6. There are up to 286 literal/length codes.  Code 256 represents the
 80       end-of-block.  Note however that the static length tree defines
 81       288 codes just to fill out the Huffman codes.  Codes 286 and 287
 82       cannot be used though, since there is no length base or extra bits
 83       defined for them.  Similarly, there are up to 30 distance codes.
 84       However, static trees define 32 codes (all 5 bits) to fill out the
 85       Huffman codes, but the last two had better not show up in the data.
 86    7. Unzip can check dynamic Huffman blocks for complete code sets.
 87       The exception is that a single code would not be complete (see #4).
 88    8. The five bits following the block type is really the number of
 89       literal codes sent minus 257.
 90    9. Length codes 8,16,16 are interpreted as 13 length codes of 8 bits
 91       (1+6+6).  Therefore, to output three times the length, you output
 92       three codes (1+1+1), whereas to output four times the same length,
 93       you only need two codes (1+3).  Hmm.
 94   10. In the tree reconstruction algorithm, Code = Code + Increment
 95       only if BitLength(i) is not zero.  (Pretty obvious.)
 96   11. Correction: 4 Bits: # of Bit Length codes - 4     (4 - 19)
 97   12. Note: length code 284 can represent 227-258, but length code 285
 98       really is 258.  The last length deserves its own, short code
 99       since it gets used a lot in very redundant files.  The length
100       258 is special since 258 - 3 (the min match length) is 255.
101   13. The literal/length and distance code bit lengths are read as a
102       single stream of lengths.  It is possible (and advantageous) for
103       a repeat code (16, 17, or 18) to go across the boundary between
104       the two sets of lengths.
105  */
106 #include <linux/compiler.h>
107 #ifdef NO_INFLATE_MALLOC
108 #include <linux/slab.h>
109 #endif
110 
111 #ifdef RCSID
112 static char rcsid[] = "#Id: inflate.c,v 0.14 1993/06/10 13:27:04 jloup Exp #";
113 #endif
114 
115 #ifndef STATIC
116 
117 #if defined(STDC_HEADERS) || defined(HAVE_STDLIB_H)
118 #  include <sys/types.h>
119 #  include <stdlib.h>
120 #endif
121 
122 #include "gzip.h"
123 #define STATIC
124 #endif /* !STATIC */
125 
126 #ifndef INIT
127 #define INIT
128 #endif
129         
130 #define slide window
131 
132 /* Huffman code lookup table entry--this entry is four bytes for machines
133    that have 16-bit pointers (e.g. PC's in the small or medium model).
134    Valid extra bits are 0..13.  e == 15 is EOB (end of block), e == 16
135    means that v is a literal, 16 < e < 32 means that v is a pointer to
136    the next table, which codes e - 16 bits, and lastly e == 99 indicates
137    an unused code.  If a code with e == 99 is looked up, this implies an
138    error in the data. */
139 struct huft {
140   uch e;                /* number of extra bits or operation */
141   uch b;                /* number of bits in this code or subcode */
142   union {
143     ush n;              /* literal, length base, or distance base */
144     struct huft *t;     /* pointer to next level of table */
145   } v;
146 };
147 
148 
149 /* Function prototypes */
150 STATIC int INIT huft_build OF((unsigned *, unsigned, unsigned, 
151                 const ush *, const ush *, struct huft **, int *));
152 STATIC int INIT huft_free OF((struct huft *));
153 STATIC int INIT inflate_codes OF((struct huft *, struct huft *, int, int));
154 STATIC int INIT inflate_stored OF((void));
155 STATIC int INIT inflate_fixed OF((void));
156 STATIC int INIT inflate_dynamic OF((void));
157 STATIC int INIT inflate_block OF((int *));
158 STATIC int INIT inflate OF((void));
159 
160 
161 /* The inflate algorithm uses a sliding 32 K byte window on the uncompressed
162    stream to find repeated byte strings.  This is implemented here as a
163    circular buffer.  The index is updated simply by incrementing and then
164    ANDing with 0x7fff (32K-1). */
165 /* It is left to other modules to supply the 32 K area.  It is assumed
166    to be usable as if it were declared "uch slide[32768];" or as just
167    "uch *slide;" and then malloc'ed in the latter case.  The definition
168    must be in unzip.h, included above. */
169 /* unsigned wp;             current position in slide */
170 #define wp outcnt
171 #define flush_output(w) (wp=(w),flush_window())
172 
173 /* Tables for deflate from PKZIP's appnote.txt. */
174 static const unsigned border[] = {    /* Order of the bit length code lengths */
175         16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15};
176 static const ush cplens[] = {         /* Copy lengths for literal codes 257..285 */
177         3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31,
178         35, 43, 51, 59, 67, 83, 99, 115, 131, 163, 195, 227, 258, 0, 0};
179         /* note: see note #13 above about the 258 in this list. */
180 static const ush cplext[] = {         /* Extra bits for literal codes 257..285 */
181         0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2,
182         3, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 0, 99, 99}; /* 99==invalid */
183 static const ush cpdist[] = {         /* Copy offsets for distance codes 0..29 */
184         1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193,
185         257, 385, 513, 769, 1025, 1537, 2049, 3073, 4097, 6145,
186         8193, 12289, 16385, 24577};
187 static const ush cpdext[] = {         /* Extra bits for distance codes */
188         0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6,
189         7, 7, 8, 8, 9, 9, 10, 10, 11, 11,
190         12, 12, 13, 13};
191 
192 
193 
194 /* Macros for inflate() bit peeking and grabbing.
195    The usage is:
196    
197         NEEDBITS(j)
198         x = b & mask_bits[j];
199         DUMPBITS(j)
200 
201    where NEEDBITS makes sure that b has at least j bits in it, and
202    DUMPBITS removes the bits from b.  The macros use the variable k
203    for the number of bits in b.  Normally, b and k are register
204    variables for speed, and are initialized at the beginning of a
205    routine that uses these macros from a global bit buffer and count.
206 
207    If we assume that EOB will be the longest code, then we will never
208    ask for bits with NEEDBITS that are beyond the end of the stream.
209    So, NEEDBITS should not read any more bytes than are needed to
210    meet the request.  Then no bytes need to be "returned" to the buffer
211    at the end of the last block.
212 
213    However, this assumption is not true for fixed blocks--the EOB code
214    is 7 bits, but the other literal/length codes can be 8 or 9 bits.
215    (The EOB code is shorter than other codes because fixed blocks are
216    generally short.  So, while a block always has an EOB, many other
217    literal/length codes have a significantly lower probability of
218    showing up at all.)  However, by making the first table have a
219    lookup of seven bits, the EOB code will be found in that first
220    lookup, and so will not require that too many bits be pulled from
221    the stream.
222  */
223 
224 STATIC ulg bb;                         /* bit buffer */
225 STATIC unsigned bk;                    /* bits in bit buffer */
226 
227 STATIC const ush mask_bits[] = {
228     0x0000,
229     0x0001, 0x0003, 0x0007, 0x000f, 0x001f, 0x003f, 0x007f, 0x00ff,
230     0x01ff, 0x03ff, 0x07ff, 0x0fff, 0x1fff, 0x3fff, 0x7fff, 0xffff
231 };
232 
233 #define NEXTBYTE()  ({ int v = get_byte(); if (v < 0) goto underrun; (uch)v; })
234 #define NEEDBITS(n) {while(k<(n)){b|=((ulg)NEXTBYTE())<<k;k+=8;}}
235 #define DUMPBITS(n) {b>>=(n);k-=(n);}
236 
237 #ifndef NO_INFLATE_MALLOC
238 /* A trivial malloc implementation, adapted from
239  *  malloc by Hannu Savolainen 1993 and Matthias Urlichs 1994
240  */
241 
242 static unsigned long malloc_ptr;
243 static int malloc_count;
244 
245 static void *malloc(int size)
246 {
247        void *p;
248 
249        if (size < 0)
250                 error("Malloc error");
251        if (!malloc_ptr)
252                 malloc_ptr = free_mem_ptr;
253 
254        malloc_ptr = (malloc_ptr + 3) & ~3;     /* Align */
255 
256        p = (void *)malloc_ptr;
257        malloc_ptr += size;
258 
259        if (free_mem_end_ptr && malloc_ptr >= free_mem_end_ptr)
260                 error("Out of memory");
261 
262        malloc_count++;
263        return p;
264 }
265 
266 static void free(void *where)
267 {
268        malloc_count--;
269        if (!malloc_count)
270                 malloc_ptr = free_mem_ptr;
271 }
272 #else
273 #define malloc(a) kmalloc(a, GFP_KERNEL)
274 #define free(a) kfree(a)
275 #endif
276 
277 /*
278    Huffman code decoding is performed using a multi-level table lookup.
279    The fastest way to decode is to simply build a lookup table whose
280    size is determined by the longest code.  However, the time it takes
281    to build this table can also be a factor if the data being decoded
282    is not very long.  The most common codes are necessarily the
283    shortest codes, so those codes dominate the decoding time, and hence
284    the speed.  The idea is you can have a shorter table that decodes the
285    shorter, more probable codes, and then point to subsidiary tables for
286    the longer codes.  The time it costs to decode the longer codes is
287    then traded against the time it takes to make longer tables.
288 
289    This results of this trade are in the variables lbits and dbits
290    below.  lbits is the number of bits the first level table for literal/
291    length codes can decode in one step, and dbits is the same thing for
292    the distance codes.  Subsequent tables are also less than or equal to
293    those sizes.  These values may be adjusted either when all of the
294    codes are shorter than that, in which case the longest code length in
295    bits is used, or when the shortest code is *longer* than the requested
296    table size, in which case the length of the shortest code in bits is
297    used.
298 
299    There are two different values for the two tables, since they code a
300    different number of possibilities each.  The literal/length table
301    codes 286 possible values, or in a flat code, a little over eight
302    bits.  The distance table codes 30 possible values, or a little less
303    than five bits, flat.  The optimum values for speed end up being
304    about one bit more than those, so lbits is 8+1 and dbits is 5+1.
305    The optimum values may differ though from machine to machine, and
306    possibly even between compilers.  Your mileage may vary.
307  */
308 
309 
310 STATIC const int lbits = 9;          /* bits in base literal/length lookup table */
311 STATIC const int dbits = 6;          /* bits in base distance lookup table */
312 
313 
314 /* If BMAX needs to be larger than 16, then h and x[] should be ulg. */
315 #define BMAX 16         /* maximum bit length of any code (16 for explode) */
316 #define N_MAX 288       /* maximum number of codes in any set */
317 
318 
319 STATIC unsigned hufts;         /* track memory usage */
320 
321 
322 STATIC int INIT huft_build(
323         unsigned *b,            /* code lengths in bits (all assumed <= BMAX) */
324         unsigned n,             /* number of codes (assumed <= N_MAX) */
325         unsigned s,             /* number of simple-valued codes (0..s-1) */
326         const ush *d,           /* list of base values for non-simple codes */
327         const ush *e,           /* list of extra bits for non-simple codes */
328         struct huft **t,        /* result: starting table */
329         int *m                  /* maximum lookup bits, returns actual */
330         )
331 /* Given a list of code lengths and a maximum table size, make a set of
332    tables to decode that set of codes.  Return zero on success, one if
333    the given code set is incomplete (the tables are still built in this
334    case), two if the input is invalid (all zero length codes or an
335    oversubscribed set of lengths), and three if not enough memory. */
336 {
337   unsigned a;                   /* counter for codes of length k */
338   unsigned f;                   /* i repeats in table every f entries */
339   int g;                        /* maximum code length */
340   int h;                        /* table level */
341   register unsigned i;          /* counter, current code */
342   register unsigned j;          /* counter */
343   register int k;               /* number of bits in current code */
344   int l;                        /* bits per table (returned in m) */
345   register unsigned *p;         /* pointer into c[], b[], or v[] */
346   register struct huft *q;      /* points to current table */
347   struct huft r;                /* table entry for structure assignment */
348   register int w;               /* bits before this table == (l * h) */
349   unsigned *xp;                 /* pointer into x */
350   int y;                        /* number of dummy codes added */
351   unsigned z;                   /* number of entries in current table */
352   struct {
353     unsigned c[BMAX+1];           /* bit length count table */
354     struct huft *u[BMAX];         /* table stack */
355     unsigned v[N_MAX];            /* values in order of bit length */
356     unsigned x[BMAX+1];           /* bit offsets, then code stack */
357   } *stk;
358   unsigned *c, *v, *x;
359   struct huft **u;
360   int ret;
361 
362 DEBG("huft1 ");
363 
364   stk = malloc(sizeof(*stk));
365   if (stk == NULL)
366     return 3;                   /* out of memory */
367 
368   c = stk->c;
369   v = stk->v;
370   x = stk->x;
371   u = stk->u;
372 
373   /* Generate counts for each bit length */
374   memzero(stk->c, sizeof(stk->c));
375   p = b;  i = n;
376   do {
377     Tracecv(*p, (stderr, (n-i >= ' ' && n-i <= '~' ? "%c %d\n" : "0x%x %d\n"), 
378             n-i, *p));
379     c[*p]++;                    /* assume all entries <= BMAX */
380     p++;                      /* Can't combine with above line (Solaris bug) */
381   } while (--i);
382   if (c[0] == n)                /* null input--all zero length codes */
383   {
384     *t = (struct huft *)NULL;
385     *m = 0;
386     ret = 2;
387     goto out;
388   }
389 
390 DEBG("huft2 ");
391 
392   /* Find minimum and maximum length, bound *m by those */
393   l = *m;
394   for (j = 1; j <= BMAX; j++)
395     if (c[j])
396       break;
397   k = j;                        /* minimum code length */
398   if ((unsigned)l < j)
399     l = j;
400   for (i = BMAX; i; i--)
401     if (c[i])
402       break;
403   g = i;                        /* maximum code length */
404   if ((unsigned)l > i)
405     l = i;
406   *m = l;
407 
408 DEBG("huft3 ");
409 
410   /* Adjust last length count to fill out codes, if needed */
411   for (y = 1 << j; j < i; j++, y <<= 1)
412     if ((y -= c[j]) < 0) {
413       ret = 2;                 /* bad input: more codes than bits */
414       goto out;
415     }
416   if ((y -= c[i]) < 0) {
417     ret = 2;
418     goto out;
419   }
420   c[i] += y;
421 
422 DEBG("huft4 ");
423 
424   /* Generate starting offsets into the value table for each length */
425   x[1] = j = 0;
426   p = c + 1;  xp = x + 2;
427   while (--i) {                 /* note that i == g from above */
428     *xp++ = (j += *p++);
429   }
430 
431 DEBG("huft5 ");
432 
433   /* Make a table of values in order of bit lengths */
434   p = b;  i = 0;
435   do {
436     if ((j = *p++) != 0)
437       v[x[j]++] = i;
438   } while (++i < n);
439   n = x[g];                   /* set n to length of v */
440 
441 DEBG("h6 ");
442 
443   /* Generate the Huffman codes and for each, make the table entries */
444   x[0] = i = 0;                 /* first Huffman code is zero */
445   p = v;                        /* grab values in bit order */
446   h = -1;                       /* no tables yet--level -1 */
447   w = -l;                       /* bits decoded == (l * h) */
448   u[0] = (struct huft *)NULL;   /* just to keep compilers happy */
449   q = (struct huft *)NULL;      /* ditto */
450   z = 0;                        /* ditto */
451 DEBG("h6a ");
452 
453   /* go through the bit lengths (k already is bits in shortest code) */
454   for (; k <= g; k++)
455   {
456 DEBG("h6b ");
457     a = c[k];
458     while (a--)
459     {
460 DEBG("h6b1 ");
461       /* here i is the Huffman code of length k bits for value *p */
462       /* make tables up to required level */
463       while (k > w + l)
464       {
465 DEBG1("1 ");
466         h++;
467         w += l;                 /* previous table always l bits */
468 
469         /* compute minimum size table less than or equal to l bits */
470         z = (z = g - w) > (unsigned)l ? l : z;  /* upper limit on table size */
471         if ((f = 1 << (j = k - w)) > a + 1)     /* try a k-w bit table */
472         {                       /* too few codes for k-w bit table */
473 DEBG1("2 ");
474           f -= a + 1;           /* deduct codes from patterns left */
475           xp = c + k;
476           if (j < z)
477             while (++j < z)       /* try smaller tables up to z bits */
478             {
479               if ((f <<= 1) <= *++xp)
480                 break;            /* enough codes to use up j bits */
481               f -= *xp;           /* else deduct codes from patterns */
482             }
483         }
484 DEBG1("3 ");
485         z = 1 << j;             /* table entries for j-bit table */
486 
487         /* allocate and link in new table */
488         if ((q = (struct huft *)malloc((z + 1)*sizeof(struct huft))) ==
489             (struct huft *)NULL)
490         {
491           if (h)
492             huft_free(u[0]);
493           ret = 3;             /* not enough memory */
494           goto out;
495         }
496 DEBG1("4 ");
497         hufts += z + 1;         /* track memory usage */
498         *t = q + 1;             /* link to list for huft_free() */
499         *(t = &(q->v.t)) = (struct huft *)NULL;
500         u[h] = ++q;             /* table starts after link */
501 
502 DEBG1("5 ");
503         /* connect to last table, if there is one */
504         if (h)
505         {
506           x[h] = i;             /* save pattern for backing up */
507           r.b = (uch)l;         /* bits to dump before this table */
508           r.e = (uch)(16 + j);  /* bits in this table */
509           r.v.t = q;            /* pointer to this table */
510           j = i >> (w - l);     /* (get around Turbo C bug) */
511           u[h-1][j] = r;        /* connect to last table */
512         }
513 DEBG1("6 ");
514       }
515 DEBG("h6c ");
516 
517       /* set up table entry in r */
518       r.b = (uch)(k - w);
519       if (p >= v + n)
520         r.e = 99;               /* out of values--invalid code */
521       else if (*p < s)
522       {
523         r.e = (uch)(*p < 256 ? 16 : 15);    /* 256 is end-of-block code */
524         r.v.n = (ush)(*p);             /* simple code is just the value */
525         p++;                           /* one compiler does not like *p++ */
526       }
527       else
528       {
529         r.e = (uch)e[*p - s];   /* non-simple--look up in lists */
530         r.v.n = d[*p++ - s];
531       }
532 DEBG("h6d ");
533 
534       /* fill code-like entries with r */
535       f = 1 << (k - w);
536       for (j = i >> w; j < z; j += f)
537         q[j] = r;
538 
539       /* backwards increment the k-bit code i */
540       for (j = 1 << (k - 1); i & j; j >>= 1)
541         i ^= j;
542       i ^= j;
543 
544       /* backup over finished tables */
545       while ((i & ((1 << w) - 1)) != x[h])
546       {
547         h--;                    /* don't need to update q */
548         w -= l;
549       }
550 DEBG("h6e ");
551     }
552 DEBG("h6f ");
553   }
554 
555 DEBG("huft7 ");
556 
557   /* Return true (1) if we were given an incomplete table */
558   ret = y != 0 && g != 1;
559 
560   out:
561   free(stk);
562   return ret;
563 }
564 
565 
566 
567 STATIC int INIT huft_free(
568         struct huft *t         /* table to free */
569         )
570 /* Free the malloc'ed tables built by huft_build(), which makes a linked
571    list of the tables it made, with the links in a dummy first entry of
572    each table. */
573 {
574   register struct huft *p, *q;
575 
576 
577   /* Go through linked list, freeing from the malloced (t[-1]) address. */
578   p = t;
579   while (p != (struct huft *)NULL)
580   {
581     q = (--p)->v.t;
582     free((char*)p);
583     p = q;
584   } 
585   return 0;
586 }
587 
588 
589 STATIC int INIT inflate_codes(
590         struct huft *tl,    /* literal/length decoder tables */
591         struct huft *td,    /* distance decoder tables */
592         int bl,             /* number of bits decoded by tl[] */
593         int bd              /* number of bits decoded by td[] */
594         )
595 /* inflate (decompress) the codes in a deflated (compressed) block.
596    Return an error code or zero if it all goes ok. */
597 {
598   register unsigned e;  /* table entry flag/number of extra bits */
599   unsigned n, d;        /* length and index for copy */
600   unsigned w;           /* current window position */
601   struct huft *t;       /* pointer to table entry */
602   unsigned ml, md;      /* masks for bl and bd bits */
603   register ulg b;       /* bit buffer */
604   register unsigned k;  /* number of bits in bit buffer */
605 
606 
607   /* make local copies of globals */
608   b = bb;                       /* initialize bit buffer */
609   k = bk;
610   w = wp;                       /* initialize window position */
611 
612   /* inflate the coded data */
613   ml = mask_bits[bl];           /* precompute masks for speed */
614   md = mask_bits[bd];
615   for (;;)                      /* do until end of block */
616   {
617     NEEDBITS((unsigned)bl)
618     if ((e = (t = tl + ((unsigned)b & ml))->e) > 16)
619       do {
620         if (e == 99)
621           return 1;
622         DUMPBITS(t->b)
623         e -= 16;
624         NEEDBITS(e)
625       } while ((e = (t = t->v.t + ((unsigned)b & mask_bits[e]))->e) > 16);
626     DUMPBITS(t->b)
627     if (e == 16)                /* then it's a literal */
628     {
629       slide[w++] = (uch)t->v.n;
630       Tracevv((stderr, "%c", slide[w-1]));
631       if (w == WSIZE)
632       {
633         flush_output(w);
634         w = 0;
635       }
636     }
637     else                        /* it's an EOB or a length */
638     {
639       /* exit if end of block */
640       if (e == 15)
641         break;
642 
643       /* get length of block to copy */
644       NEEDBITS(e)
645       n = t->v.n + ((unsigned)b & mask_bits[e]);
646       DUMPBITS(e);
647 
648       /* decode distance of block to copy */
649       NEEDBITS((unsigned)bd)
650       if ((e = (t = td + ((unsigned)b & md))->e) > 16)
651         do {
652           if (e == 99)
653             return 1;
654           DUMPBITS(t->b)
655           e -= 16;
656           NEEDBITS(e)
657         } while ((e = (t = t->v.t + ((unsigned)b & mask_bits[e]))->e) > 16);
658       DUMPBITS(t->b)
659       NEEDBITS(e)
660       d = w - t->v.n - ((unsigned)b & mask_bits[e]);
661       DUMPBITS(e)
662       Tracevv((stderr,"\\[%d,%d]", w-d, n));
663 
664       /* do the copy */
665       do {
666         n -= (e = (e = WSIZE - ((d &= WSIZE-1) > w ? d : w)) > n ? n : e);
667 #if !defined(NOMEMCPY) && !defined(DEBUG)
668         if (w - d >= e)         /* (this test assumes unsigned comparison) */
669         {
670           memcpy(slide + w, slide + d, e);
671           w += e;
672           d += e;
673         }
674         else                      /* do it slow to avoid memcpy() overlap */
675 #endif /* !NOMEMCPY */
676           do {
677             slide[w++] = slide[d++];
678             Tracevv((stderr, "%c", slide[w-1]));
679           } while (--e);
680         if (w == WSIZE)
681         {
682           flush_output(w);
683           w = 0;
684         }
685       } while (n);
686     }
687   }
688 
689 
690   /* restore the globals from the locals */
691   wp = w;                       /* restore global window pointer */
692   bb = b;                       /* restore global bit buffer */
693   bk = k;
694 
695   /* done */
696   return 0;
697 
698  underrun:
699   return 4;                     /* Input underrun */
700 }
701 
702 
703 
704 STATIC int INIT inflate_stored(void)
705 /* "decompress" an inflated type 0 (stored) block. */
706 {
707   unsigned n;           /* number of bytes in block */
708   unsigned w;           /* current window position */
709   register ulg b;       /* bit buffer */
710   register unsigned k;  /* number of bits in bit buffer */
711 
712 DEBG("<stor");
713 
714   /* make local copies of globals */
715   b = bb;                       /* initialize bit buffer */
716   k = bk;
717   w = wp;                       /* initialize window position */
718 
719 
720   /* go to byte boundary */
721   n = k & 7;
722   DUMPBITS(n);
723 
724 
725   /* get the length and its complement */
726   NEEDBITS(16)
727   n = ((unsigned)b & 0xffff);
728   DUMPBITS(16)
729   NEEDBITS(16)
730   if (n != (unsigned)((~b) & 0xffff))
731     return 1;                   /* error in compressed data */
732   DUMPBITS(16)
733 
734 
735   /* read and output the compressed data */
736   while (n--)
737   {
738     NEEDBITS(8)
739     slide[w++] = (uch)b;
740     if (w == WSIZE)
741     {
742       flush_output(w);
743       w = 0;
744     }
745     DUMPBITS(8)
746   }
747 
748 
749   /* restore the globals from the locals */
750   wp = w;                       /* restore global window pointer */
751   bb = b;                       /* restore global bit buffer */
752   bk = k;
753 
754   DEBG(">");
755   return 0;
756 
757  underrun:
758   return 4;                     /* Input underrun */
759 }
760 
761 
762 /*
763  * We use `noinline' here to prevent gcc-3.5 from using too much stack space
764  */
765 STATIC int noinline INIT inflate_fixed(void)
766 /* decompress an inflated type 1 (fixed Huffman codes) block.  We should
767    either replace this with a custom decoder, or at least precompute the
768    Huffman tables. */
769 {
770   int i;                /* temporary variable */
771   struct huft *tl;      /* literal/length code table */
772   struct huft *td;      /* distance code table */
773   int bl;               /* lookup bits for tl */
774   int bd;               /* lookup bits for td */
775   unsigned *l;          /* length list for huft_build */
776 
777 DEBG("<fix");
778 
779   l = malloc(sizeof(*l) * 288);
780   if (l == NULL)
781     return 3;                   /* out of memory */
782 
783   /* set up literal table */
784   for (i = 0; i < 144; i++)
785     l[i] = 8;
786   for (; i < 256; i++)
787     l[i] = 9;
788   for (; i < 280; i++)
789     l[i] = 7;
790   for (; i < 288; i++)          /* make a complete, but wrong code set */
791     l[i] = 8;
792   bl = 7;
793   if ((i = huft_build(l, 288, 257, cplens, cplext, &tl, &bl)) != 0) {
794     free(l);
795     return i;
796   }
797 
798   /* set up distance table */
799   for (i = 0; i < 30; i++)      /* make an incomplete code set */
800     l[i] = 5;
801   bd = 5;
802   if ((i = huft_build(l, 30, 0, cpdist, cpdext, &td, &bd)) > 1)
803   {
804     huft_free(tl);
805     free(l);
806 
807     DEBG(">");
808     return i;
809   }
810 
811 
812   /* decompress until an end-of-block code */
813   if (inflate_codes(tl, td, bl, bd)) {
814     free(l);
815     return 1;
816   }
817 
818   /* free the decoding tables, return */
819   free(l);
820   huft_free(tl);
821   huft_free(td);
822   return 0;
823 }
824 
825 
826 /*
827  * We use `noinline' here to prevent gcc-3.5 from using too much stack space
828  */
829 STATIC int noinline INIT inflate_dynamic(void)
830 /* decompress an inflated type 2 (dynamic Huffman codes) block. */
831 {
832   int i;                /* temporary variables */
833   unsigned j;
834   unsigned l;           /* last length */
835   unsigned m;           /* mask for bit lengths table */
836   unsigned n;           /* number of lengths to get */
837   struct huft *tl;      /* literal/length code table */
838   struct huft *td;      /* distance code table */
839   int bl;               /* lookup bits for tl */
840   int bd;               /* lookup bits for td */
841   unsigned nb;          /* number of bit length codes */
842   unsigned nl;          /* number of literal/length codes */
843   unsigned nd;          /* number of distance codes */
844   unsigned *ll;         /* literal/length and distance code lengths */
845   register ulg b;       /* bit buffer */
846   register unsigned k;  /* number of bits in bit buffer */
847   int ret;
848 
849 DEBG("<dyn");
850 
851 #ifdef PKZIP_BUG_WORKAROUND
852   ll = malloc(sizeof(*ll) * (288+32));  /* literal/length and distance code lengths */
853 #else
854   ll = malloc(sizeof(*ll) * (286+30));  /* literal/length and distance code lengths */
855 #endif
856 
857   if (ll == NULL)
858     return 1;
859 
860   /* make local bit buffer */
861   b = bb;
862   k = bk;
863 
864 
865   /* read in table lengths */
866   NEEDBITS(5)
867   nl = 257 + ((unsigned)b & 0x1f);      /* number of literal/length codes */
868   DUMPBITS(5)
869   NEEDBITS(5)
870   nd = 1 + ((unsigned)b & 0x1f);        /* number of distance codes */
871   DUMPBITS(5)
872   NEEDBITS(4)
873   nb = 4 + ((unsigned)b & 0xf);         /* number of bit length codes */
874   DUMPBITS(4)
875 #ifdef PKZIP_BUG_WORKAROUND
876   if (nl > 288 || nd > 32)
877 #else
878   if (nl > 286 || nd > 30)
879 #endif
880   {
881     ret = 1;             /* bad lengths */
882     goto out;
883   }
884 
885 DEBG("dyn1 ");
886 
887   /* read in bit-length-code lengths */
888   for (j = 0; j < nb; j++)
889   {
890     NEEDBITS(3)
891     ll[border[j]] = (unsigned)b & 7;
892     DUMPBITS(3)
893   }
894   for (; j < 19; j++)
895     ll[border[j]] = 0;
896 
897 DEBG("dyn2 ");
898 
899   /* build decoding table for trees--single level, 7 bit lookup */
900   bl = 7;
901   if ((i = huft_build(ll, 19, 19, NULL, NULL, &tl, &bl)) != 0)
902   {
903     if (i == 1)
904       huft_free(tl);
905     ret = i;                   /* incomplete code set */
906     goto out;
907   }
908 
909 DEBG("dyn3 ");
910 
911   /* read in literal and distance code lengths */
912   n = nl + nd;
913   m = mask_bits[bl];
914   i = l = 0;
915   while ((unsigned)i < n)
916   {
917     NEEDBITS((unsigned)bl)
918     j = (td = tl + ((unsigned)b & m))->b;
919     DUMPBITS(j)
920     j = td->v.n;
921     if (j < 16)                 /* length of code in bits (0..15) */
922       ll[i++] = l = j;          /* save last length in l */
923     else if (j == 16)           /* repeat last length 3 to 6 times */
924     {
925       NEEDBITS(2)
926       j = 3 + ((unsigned)b & 3);
927       DUMPBITS(2)
928       if ((unsigned)i + j > n) {
929         ret = 1;
930         goto out;
931       }
932       while (j--)
933         ll[i++] = l;
934     }
935     else if (j == 17)           /* 3 to 10 zero length codes */
936     {
937       NEEDBITS(3)
938       j = 3 + ((unsigned)b & 7);
939       DUMPBITS(3)
940       if ((unsigned)i + j > n) {
941         ret = 1;
942         goto out;
943       }
944       while (j--)
945         ll[i++] = 0;
946       l = 0;
947     }
948     else                        /* j == 18: 11 to 138 zero length codes */
949     {
950       NEEDBITS(7)
951       j = 11 + ((unsigned)b & 0x7f);
952       DUMPBITS(7)
953       if ((unsigned)i + j > n) {
954         ret = 1;
955         goto out;
956       }
957       while (j--)
958         ll[i++] = 0;
959       l = 0;
960     }
961   }
962 
963 DEBG("dyn4 ");
964 
965   /* free decoding table for trees */
966   huft_free(tl);
967 
968 DEBG("dyn5 ");
969 
970   /* restore the global bit buffer */
971   bb = b;
972   bk = k;
973 
974 DEBG("dyn5a ");
975 
976   /* build the decoding tables for literal/length and distance codes */
977   bl = lbits;
978   if ((i = huft_build(ll, nl, 257, cplens, cplext, &tl, &bl)) != 0)
979   {
980 DEBG("dyn5b ");
981     if (i == 1) {
982       error("incomplete literal tree");
983       huft_free(tl);
984     }
985     ret = i;                   /* incomplete code set */
986     goto out;
987   }
988 DEBG("dyn5c ");
989   bd = dbits;
990   if ((i = huft_build(ll + nl, nd, 0, cpdist, cpdext, &td, &bd)) != 0)
991   {
992 DEBG("dyn5d ");
993     if (i == 1) {
994       error("incomplete distance tree");
995 #ifdef PKZIP_BUG_WORKAROUND
996       i = 0;
997     }
998 #else
999       huft_free(td);
1000     }
1001     huft_free(tl);
1002     ret = i;                   /* incomplete code set */
1003     goto out;
1004 #endif
1005   }
1006 
1007 DEBG("dyn6 ");
1008 
1009   /* decompress until an end-of-block code */
1010   if (inflate_codes(tl, td, bl, bd)) {
1011     ret = 1;
1012     goto out;
1013   }
1014 
1015 DEBG("dyn7 ");
1016 
1017   /* free the decoding tables, return */
1018   huft_free(tl);
1019   huft_free(td);
1020 
1021   DEBG(">");
1022   ret = 0;
1023 out:
1024   free(ll);
1025   return ret;
1026 
1027 underrun:
1028   ret = 4;                      /* Input underrun */
1029   goto out;
1030 }
1031 
1032 
1033 
1034 STATIC int INIT inflate_block(
1035         int *e                  /* last block flag */
1036         )
1037 /* decompress an inflated block */
1038 {
1039   unsigned t;           /* block type */
1040   register ulg b;       /* bit buffer */
1041   register unsigned k;  /* number of bits in bit buffer */
1042 
1043   DEBG("<blk");
1044 
1045   /* make local bit buffer */
1046   b = bb;
1047   k = bk;
1048 
1049 
1050   /* read in last block bit */
1051   NEEDBITS(1)
1052   *e = (int)b & 1;
1053   DUMPBITS(1)
1054 
1055 
1056   /* read in block type */
1057   NEEDBITS(2)
1058   t = (unsigned)b & 3;
1059   DUMPBITS(2)
1060 
1061 
1062   /* restore the global bit buffer */
1063   bb = b;
1064   bk = k;
1065 
1066   /* inflate that block type */
1067   if (t == 2)
1068     return inflate_dynamic();
1069   if (t == 0)
1070     return inflate_stored();
1071   if (t == 1)
1072     return inflate_fixed();
1073 
1074   DEBG(">");
1075 
1076   /* bad block type */
1077   return 2;
1078 
1079  underrun:
1080   return 4;                     /* Input underrun */
1081 }
1082 
1083 
1084 
1085 STATIC int INIT inflate(void)
1086 /* decompress an inflated entry */
1087 {
1088   int e;                /* last block flag */
1089   int r;                /* result code */
1090   unsigned h;           /* maximum struct huft's malloc'ed */
1091 
1092   /* initialize window, bit buffer */
1093   wp = 0;
1094   bk = 0;
1095   bb = 0;
1096 
1097 
1098   /* decompress until the last block */
1099   h = 0;
1100   do {
1101     hufts = 0;
1102 #ifdef ARCH_HAS_DECOMP_WDOG
1103     arch_decomp_wdog();
1104 #endif
1105     r = inflate_block(&e);
1106     if (r)
1107             return r;
1108     if (hufts > h)
1109       h = hufts;
1110   } while (!e);
1111 
1112   /* Undo too much lookahead. The next read will be byte aligned so we
1113    * can discard unused bits in the last meaningful byte.
1114    */
1115   while (bk >= 8) {
1116     bk -= 8;
1117     inptr--;
1118   }
1119 
1120   /* flush out slide */
1121   flush_output(wp);
1122 
1123 
1124   /* return success */
1125 #ifdef DEBUG
1126   fprintf(stderr, "<%u> ", h);
1127 #endif /* DEBUG */
1128   return 0;
1129 }
1130 
1131 /**********************************************************************
1132  *
1133  * The following are support routines for inflate.c
1134  *
1135  **********************************************************************/
1136 
1137 static ulg crc_32_tab[256];
1138 static ulg crc;         /* initialized in makecrc() so it'll reside in bss */
1139 #define CRC_VALUE (crc ^ 0xffffffffUL)
1140 
1141 /*
1142  * Code to compute the CRC-32 table. Borrowed from 
1143  * gzip-1.0.3/makecrc.c.
1144  */
1145 
1146 static void INIT
1147 makecrc(void)
1148 {
1149 /* Not copyrighted 1990 Mark Adler      */
1150 
1151   unsigned long c;      /* crc shift register */
1152   unsigned long e;      /* polynomial exclusive-or pattern */
1153   int i;                /* counter for all possible eight bit values */
1154   int k;                /* byte being shifted into crc apparatus */
1155 
1156   /* terms of polynomial defining this crc (except x^32): */
1157   static const int p[] = {0,1,2,4,5,7,8,10,11,12,16,22,23,26};
1158 
1159   /* Make exclusive-or pattern from polynomial */
1160   e = 0;
1161   for (i = 0; i < sizeof(p)/sizeof(int); i++)
1162     e |= 1L << (31 - p[i]);
1163 
1164   crc_32_tab[0] = 0;
1165 
1166   for (i = 1; i < 256; i++)
1167   {
1168     c = 0;
1169     for (k = i | 256; k != 1; k >>= 1)
1170     {
1171       c = c & 1 ? (c >> 1) ^ e : c >> 1;
1172       if (k & 1)
1173         c ^= e;
1174     }
1175     crc_32_tab[i] = c;
1176   }
1177 
1178   /* this is initialized here so this code could reside in ROM */
1179   crc = (ulg)0xffffffffUL; /* shift register contents */
1180 }
1181 
1182 /* gzip flag byte */
1183 #define ASCII_FLAG   0x01 /* bit 0 set: file probably ASCII text */
1184 #define CONTINUATION 0x02 /* bit 1 set: continuation of multi-part gzip file */
1185 #define EXTRA_FIELD  0x04 /* bit 2 set: extra field present */
1186 #define ORIG_NAME    0x08 /* bit 3 set: original file name present */
1187 #define COMMENT      0x10 /* bit 4 set: file comment present */
1188 #define ENCRYPTED    0x20 /* bit 5 set: file is encrypted */
1189 #define RESERVED     0xC0 /* bit 6,7:   reserved */
1190 
1191 /*
1192  * Do the uncompression!
1193  */
1194 static int INIT gunzip(void)
1195 {
1196     uch flags;
1197     unsigned char magic[2]; /* magic header */
1198     char method;
1199     ulg orig_crc = 0;       /* original crc */
1200     ulg orig_len = 0;       /* original uncompressed length */
1201     int res;
1202 
1203     magic[0] = NEXTBYTE();
1204     magic[1] = NEXTBYTE();
1205     method   = NEXTBYTE();
1206 
1207     if (magic[0] != 037 ||
1208         ((magic[1] != 0213) && (magic[1] != 0236))) {
1209             error("bad gzip magic numbers");
1210             return -1;
1211     }
1212 
1213     /* We only support method #8, DEFLATED */
1214     if (method != 8)  {
1215             error("internal error, invalid method");
1216             return -1;
1217     }
1218 
1219     flags  = (uch)get_byte();
1220     if ((flags & ENCRYPTED) != 0) {
1221             error("Input is encrypted");
1222             return -1;
1223     }
1224     if ((flags & CONTINUATION) != 0) {
1225             error("Multi part input");
1226             return -1;
1227     }
1228     if ((flags & RESERVED) != 0) {
1229             error("Input has invalid flags");
1230             return -1;
1231     }
1232     NEXTBYTE(); /* Get timestamp */
1233     NEXTBYTE();
1234     NEXTBYTE();
1235     NEXTBYTE();
1236 
1237     (void)NEXTBYTE();  /* Ignore extra flags for the moment */
1238     (void)NEXTBYTE();  /* Ignore OS type for the moment */
1239 
1240     if ((flags & EXTRA_FIELD) != 0) {
1241             unsigned len = (unsigned)NEXTBYTE();
1242             len |= ((unsigned)NEXTBYTE())<<8;
1243             while (len--) (void)NEXTBYTE();
1244     }
1245 
1246     /* Get original file name if it was truncated */
1247     if ((flags & ORIG_NAME) != 0) {
1248             /* Discard the old name */
1249             while (NEXTBYTE() != 0) /* null */ ;
1250     } 
1251 
1252     /* Discard file comment if any */
1253     if ((flags & COMMENT) != 0) {
1254             while (NEXTBYTE() != 0) /* null */ ;
1255     }
1256 
1257     /* Decompress */
1258     if ((res = inflate())) {
1259             switch (res) {
1260             case 0:
1261                     break;
1262             case 1:
1263                     error("invalid compressed format (err=1)");
1264                     break;
1265             case 2:
1266                     error("invalid compressed format (err=2)");
1267                     break;
1268             case 3:
1269                     error("out of memory");
1270                     break;
1271             case 4:
1272                     error("out of input data");
1273                     break;
1274             default:
1275                     error("invalid compressed format (other)");
1276             }
1277             return -1;
1278     }
1279             
1280     /* Get the crc and original length */
1281     /* crc32  (see algorithm.doc)
1282      * uncompressed input size modulo 2^32
1283      */
1284     orig_crc = (ulg) NEXTBYTE();
1285     orig_crc |= (ulg) NEXTBYTE() << 8;
1286     orig_crc |= (ulg) NEXTBYTE() << 16;
1287     orig_crc |= (ulg) NEXTBYTE() << 24;
1288     
1289     orig_len = (ulg) NEXTBYTE();
1290     orig_len |= (ulg) NEXTBYTE() << 8;
1291     orig_len |= (ulg) NEXTBYTE() << 16;
1292     orig_len |= (ulg) NEXTBYTE() << 24;
1293     
1294     /* Validate decompression */
1295     if (orig_crc != CRC_VALUE) {
1296             error("crc error");
1297             return -1;
1298     }
1299     if (orig_len != bytes_out) {
1300             error("length error");
1301             return -1;
1302     }
1303     return 0;
1304 
1305  underrun:                      /* NEXTBYTE() goto's here if needed */
1306     error("out of input data");
1307     return -1;
1308 }
1309 
1310 
1311 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp